Open Data Don't Just Increase Transparency - They Improve the Science
Open Data Don't Just Increase Transparency. They Improve the Science.
Launching a new AFLEARN Insights series on what open assessment data make possible
The case for open data is often framed in terms of transparency, particularly in research. Publicly funded data should be available for scrutiny; published results should be reproducible; and researchers should be able to verify the claims made from them.
All of this matters. But it captures only part of the value of open microdata.
Our new report, From Assessment Data to Better Evidence: Strengthening Africa's Foundational Learning Evidence Ecosystem, examines what happens to foundational learning data after it is collected — and argues for treating it as a long-term public asset rather than a one-off project output. This is the first article in a new AFLEARN Insights series exploring what that report found, and it makes the case that has convinced us most: open data don't just increase transparency. They improve the science itself.
Published reports cannot answer every question
Assessment reports necessarily focus on a limited set of questions. They present the indicators that matter most to governments and assessment programmes, describe major patterns and communicate findings to policymakers and the public.
Microdata allow researchers to go further. They make it possible to investigate who is represented in an assessment, how response patterns affect the realised sample, what lies beneath national averages, how inequalities vary across the learning distribution, and whether apparent differences between countries could partly reflect assessment design rather than actual differences in learning.
Secondary analysis can test assumptions
Many assumptions underlying comparative assessment are difficult to evaluate from published tables alone. Are children in the same grade being assessed at similar points in the school year? Do assessment samples represent similar school sectors? Are reading and mathematics results based on the same children? Does the language of assessment align with children's language of instruction?
When microdata and documentation are available, these become empirical questions rather than assumptions. In one of our analyses, two countries posted almost identical proficiency rates in mathematics — until we looked inside the numbers and found two completely different learning stories. We return to that example later in this series.
New evidence without new data collection
This is one of the most powerful features of secondary analysis. It can generate new evidence from investments that have already been made.
The questions explored in our report—representation, learning distributions, inequality, instructional exposure and language—were addressed through existing data, without the need to collect additional data. What they required was access to those data and the ability to interrogate them carefully.
Open data as scientific infrastructure
Transparency and reproducibility remain compelling reasons for data sharing. But we should also recognise open microdata as research infrastructure.
They allow findings to be challenged and extended, methods to improve, new questions to be asked and different datasets to be connected. In doing so, they increase both the credibility and the cumulative value of assessment programmes.
Data sharing should therefore not be treated as the final administrative step in an assessment. It is part of the scientific process itself, opening the door to further value creation and greater returns from investments in data collection.